Present invention embodiments relate to database queries, and more specifically, to identifying and discarding non-essential statistics used for query optimization.
Statistics may be used to characterize content of data stored in a database system in order to ensure good query performance. As the data stored in the database grows in size and changes in content over time, corresponding statistics may no longer accurately reflect the content of the data stored in the database.
Whether to remove a statistic is generally determined based upon knowledge of a particular query or a set of queries (a query domain). However, since the determination to remove a particular statistic is based upon known queries, subsequent new or unknown queries not falling within the query domain may lead to a substantial regression in query performance.